Close Window

Pennsylvania State University: "Targeted observations in the winter storm reconnaissance program"

Final Report


1.1 The codes used for the identification of the most sensitive observational areas, formally run on PSU workstations and on the NCEP Cray supercomputers, were ported onto the IBM SP machine. Changes were introduced that made the code compatible with operational standards. Also, the codes and scripts were optimized to shorten run times.

1.2 An improved user interface has been developed that facilitates the use of the above codes in an operational setting by the Senior Duty Meteorologists (SDM) of NCO/NCEP. User friendly input and output formats were designed, tested, and implemented in NCEP operations. In the operational Winter Storm Reconnaissance (WSR) program, the SDM personnel enters input data (time and location of a threatening weather event) by means of an interactive script, which then creates a file. This file is in turn serves as input for the sensitivity job that is initiated by the collocated IBM SP operators, per the request of the SDM. Graphical output from the sensitivity job is made available to the SDM personnel for decision making purposes in the form of printed hardcopy and web based electronic figures. Related work was done in collaboration with SDM and Senior Program Analyst (SPA) personnel to ensure user friendly and operationally compatible solutions.

1.3 Progress has been made in the area of automating the objective verification of forecasts generated in parallel analysis/forecast cycles. These cycles are run at NCEP to monitor the impact of observational data in near real time. The parallel assimilation/forecast cycle is identical to the operational cycle except all adaptively taken dropsonde data are excluded from it. The final test of the utility of the WSR program is a comparison of the operational and experimental forecasts with observational data. We inserted verification codes routinely used at EMC into the parallel scripts. Verification data can now be collected with minimal human intervention for wind, temperature and surface pressure observations.

1.4 The operationally implemented WSR targeting algorithm was designed in such a way that it can be used for mesoscale targeting as well. Beyond the WSR default, two special configurations were prepared, one for use along the west (PACJET, Pacific Jets Experiment), and another along the east coast of the US. The smaller search areas facilitate more refined sensitivity calculations, though at this stage they are still based on 2.5x2.5 lat/lon ensemble forecast data. 1x1 lat/lon ensemble data from ECMWF is not yet available.


2.1 Student visits. Jon Moskaitis, a student of Craig Bishop, visited NCEP during two extended periods. These visits not only advanced the joint research but also contributed to the professional development of the student.

2.2 Intensive collaboration between NCEP and PSU scientists. In the area of adaptive observations, interaction among scientists at the two institutions, partly due to the COMET collaboration, became more intense.


Two manuscripts are being prepared for publication in the Bulletin of American Meteorological Society, and in Weather and Forecasting. A seminar presentation is also planned at NCEP on the subject. Note that the project included intensive, one-on-one training sessions on a daily basis during the WSR program (15 January - 20 February 2001) for SDM personnel, provided by Zoltan Toth. As a result, SDM will be ready to run the extended WSR02 program after a brief refresher training session conducted before the next program starts. It is hoped that the student's honor thesis will lead to a published research contribution to the field of targeted observations. The student is currently in the process of narrowing down his research aims.


Toth, Z., I. Szunyogh, C. Bishop, S. Majumdar, R. Morss, J. Moskaitis, D. Reynolds, D. Weinbrenner, D. Michaud, N. Surgi, M. Ralph, J. Parrish, J. Talbot, J. Pavone, and S. Lord, 2002: Adaptive observations at NCEP: Past, present, and future. Preprints for the Symposium on Observations, Data Assimilation, and Probabilistic Prediction, 13-17 January 2002, Orlando, FL, in print.

Related publications:

Bishop, C. H., B. J., Etherton, and S. J. Majumdar, 2001: Adaptive Sampling with the Ensemble Transform Kalman Filter. Part I: Theoretical Aspects. Mon. Wea. Rev. 129, 420-436.

Majumdar, S.J., C.H. Bishop, B.J. Etherton, and Z. Toth, 2001: Adaptive Sampling with the Ensemble Transform Kalman Filter. Part II: Field Program Implementation. Mon. Wea. Rev., in print.

Majumdar, S.J., C.H. Bishop, B. J. Etherton, I. Szunyogh, and Z. Toth, 2001: Can an Ensemble Transform Kalman Filter predict the reduction in forecast error variance produced by targeted observations? Q. J. R. Meteor. Soc., in print.

Szunyogh, I. , Z. Toth, S. Majumdar, Zimin, A. V., and A. Persson, 2001: On the propagation of the effect of targeted observations: The 2000 Winter Storm Reconnaissance Program. Mon. Wea. Rev., in print.

Toth, Z., I. Szunyogh, C. Bishop, S. Majumdar, R. Morss, and S. Lord, 2001: On the use of targeted observations in operational numerical weather prediction. Preprints for the Fifth Symposium on Integrated Observing Systems, January, 15-19, 2001, Albuquerque, NM, AMS, 72-79.


4.1 A data base has been begun detailing times and dates when forecasts for significant east-coast weather events were highly sensitive to central and western US observations. This data base will aid studies of targeted observations.

4.2 As a by-product of the developmental work described under points 1.1 and 1.2, a script and associated codes were developed for the research use of EMC or other interested personnel. With these scripts and codes past WSR or other cases can be further analyzed.


5.1 This work has provided a unique opportunity for the student involved in the work to familiarize himself with the operational work environment and requirements within NWS. In addition he has accumulated technical knowledge and experience with respect to atmospheric dynamics, predictability, and synoptic meteorology as used in practice. The student is a member of Penn State's Schreyer honors college and will use his experience with the NCEP/PSU targeting system as the basis for some original research for his honors thesis.

5.2 After the completion of work under 1.1 and 1.2, the WSR program, with its related scripts and codes, was operationally implemented at NWS/NCEP. This was the major goal of the project and it has been successfully accomplished. This was possible only with the collaborative work COMET supported. Preliminary results from the WSR01 program indicate that approximately 70% of the forecasts that were targeted by the use of adaptive observations improved in quality due to those observations. This rate of forecast improvement is in line with results from earlier field programs and can be expected from future programs as well.


6.1 University

The participating student was not able to obtain direct access to the operational targeting code for security reasons. This made it difficult for him to provide the level of software support that he would have liked to give.

6.2 NWS

a) A key personnel working on adaptive observations left EMC. His departure slowed progress under points 1.3 and 1.4. A replacement has not been found yet.

b) The operational implementation of the software took more effort and time than anticipated. This slowed down progress that could have been achieved otherwise under points 1.3 and 1.4.

c) A bug was found and later corrected in the operational implementation of the sensitivity calculations.

d) Problems arose when attempting to read precipitation files under point 1.3. A solution has been found that will allow the inclusion of precipitation in the norm used in the targeting procedure.


* Incorporate precipitation into the group of verified variables
* Fully automate the verification algorithm in the parallel analysis/forecast cycles
* Add a new functionality to the operational targeting script that could be used by SDM to track the forecast impact of data from a selected location. This can be beneficial for quickly estimating the impact of missing or erroneous data.
* Modify or generalize the scripts/codes discussed under points 1.1 and 1.2 so they can be used with 1x1 lat/lon ensemble data for mesoscale targeting applications.